Summary

With the workflow presented here with the aid of the QGIS plugins profileAAR and georeferencer, it can be shown that a comparable result can be achieved for rectifying archaeological profile photos with the open source applications as with the proprietary software that was previously in use. Disadvantages compared to the CAD-based way become apparent when the profiles are to be viewed and processed in 3D space. The differences in the compared measuring sections are deviations of less than one centimeter for 29 profiles, which is to be regarded as sufficient for the application. The method can thus be seen as an alternative, at least for simple applications.

Statement of need / Introduction

For a long time, the standard method for documenting archaeological profiles and plana has been drawing in the field. In addition to the classic hand drawings, new, faster innovations came up in this area. This certainly included the pantograph in which an analog drawing was created or the Trigomat, which enabled direct digital drawing. With the advent of total station measurement and the connection to CAD systems, more and more recording was shifted directly to the PC. In order to include photographs of planes and profiles in these measurements, image-based methods (e.g. structure from motion) are of course available for more complex questions, but photogrammetry is also a possibility to integrate profiles and planes. In order to implement this on the software side, the AutoCAD plugin PhoToPlan (Formerly from Kubit, now Faro AsBuild Photo) has prevailed (Gütter 2015). It should be noted, however, that both AutoCAD and the plug-in are licensed and expensive to use. In the context of the increasing focus on the use of open source GIS in archeology and also in excavation technology (e.g. survey2gis or tachy2gis), the aim was to evaluate to what extent the functionality of the rectification of profiles and plans in QGIS was possible to implement. Using the georeferencer plug-in integrated in QGIS, it was previously possible to display Plana, but this was more difficult with profiles. The problem here is that the points measured in the 3D space map the profile, but QGIS can only work with the 2.5D space. The plug-in profileAAR (Mennenga, Schmütz, and Rinne 2019) was developed to transform the points in such a way that rectification in QGIS is possible. It is possible to document profiles using SFM and thus form the basis for drawings and the like. In many cases (simple post holes or sectional profiles) the effort involved is not in proportion to the result, so that the possibility of rectifying single images should still be available. In the following, the functionality of the plug-in and the rectification of profiles, but also the accuracy in comparison with the rectification in the previous approach using Faro AsBuild are presented.

profileAAR

Fig. 1: profileAAR Plugin.

Fig. 1: profileAAR Plugin.

The aim of the plug-in profileAAR (Fig. 1) was to create the possibility to transform the measured photogrammetry points in such a way that it is possible to rectify the photographs using the georeferencer integrated in QGIS. The Plugin is available on github and in the qgis extension manager (QGIS Plugin Repository 2021). To do this, they had to be moved from their position to the 2D view in such a way that they could be used. In a classic hand drawing, two nails are attached to the profile at the same height and these are connected with a cord that is used as a zero line. This line is measured vertically up and down as possible in order to obtain the support points for a drawing. This means that the profile, whether straight or crooked, is in principle projected onto a vertical plane. Theoretically, I can achieve the same thing with photography if I align the camera horizontally in a balance and vertically orthogonally on the profile. Since this is hardly feasible in practice, georeferencing is necessary. As a basis, a file is required in which the points are in a Cartesian coordinate system and the following parameters are available in columns at each point: z-value, profile number, view of the profile and whether the points should be used for calculating the rotation parameters (Tab. 1).

z-Value section_no use
0.5 1 N
0.5 1 N
0.25 1 N
0 1 N
0 1 N
1.8 2 E
1.5 2 E
1 2 E
0.75 2 E
0.8 2 E

Tab. 1: Example of an attribute table

Several steps are necessary to prepare the profile nails for use - the points must be rotated around the z-axis so that the profile is parallel to the x-axis. In a subsequent step, they must then be rotated around the X axis. For each profile it must first be determined how it can be aligned parallel to the x-axis. In order to determine the rotation parameters, the mean in the point cloud of a profile is determined using a linear regression (Backhaus et al. 2011) and the angle of rotation can then be calculated on the basis of the gradient. There are two important parameters that can be specified. On the one hand, it is important to know from which direction the photo was taken (view Fig. 1) and which points should be used to determine the rotation angle (used Fig. 1). With the latter, outliers in the measurement can be removed for the parameter determination - but they are still rotated (this procedure also allows finding points or similar to be attached to the profile, but only to rotate them and thus represent them on the profile). After determining the angle, the points are now aligned parallel to the X axis via translation and rotation. (In this context, the Direction selection menu (Fig. 1) can be used to determine whether the points are parallel to the profile section (original), parallel to the x-axis (horizontal) or parallel to the x-axis but with the smallest x-value = 0 ( original heigth). The latter enables a coordinate frame to be displayed around the profile in the layout function of Qgis.) In the next step, the profile is folded in such a way that it is no longer vertical in space, but lies on the 2D surface. In the simple case, the z and y coordinates are swapped (Fig. 1 Method - Projected). This method methodically produces the result that is also achieved when a classic drawing is carried out in the field, as described above (Fig. 2). It is also possible to carry out this step in such a way that an orthogonal view of the profile is generated (Fig. 1 Method - Surface; Fig. 2-2). The procedure is the same as for the rotation around the z-axis.

Fig. 2: Schematic representation of the various options in the method menu.

Fig. 2: Schematic representation of the various options in the method menu.

To complete the functions, there is still the option of extracting the upper right photogrammetry points in order to be able to use them in the layout as height markers and the profile section, based on the outermost, highest nails, can be extracted as a shape file. The georeferencer can now be used to finally rectify the photos of the profiles. After loading the images, the photogrammetry nails can now be picked up and connected to the rotated nails from profileAAR. In the following, the image is rectified using projective transformation (Documentation 2021).

Method for compairing the procedures

In order to compare the results of the rectification in QGIS with the previous and common method in AutoCAD, a test series of 29 profiles was rectified in both QGIS and AutoCAD. Since the processing of the images in the programs is very different and a pixel-by-pixel comparison cannot be carried out, a manual approach was therefore chosen. In order to be able to compare the procedures, the procedure had to be adapted as detailed as possible. For this purpose, a free UCS was first placed over the measured photogrammetry points in AsBuild Photo and the image on it was rectified. In the next step, a vertical UCS was created over the points and several points that were distributed and easily identified were marked in the image. The same points were then also marked in the rectified image in the GIS and the distances between the points were measured (Fig. 3). These distances form the basis of the comparison (Tab. 2).

Profile no. Axis no. Distance [cm] Qgis Distance [cm] CAD Area No. of pictures Difference [cm]
16 1 85.663 85.779 0.2 m² 1 -0.116
16 2 38.643 38.696 1.4 x 0.23 m -0.053
16 3 32.796 32.788 0.008
16 4 31.372 31.324 0.048
23 1 32.326 32.348 0.5 m² 1 -0.022
23 2 80.669 80.64 2.35 x 0.55m 0.029
23 3 153.927 153.929 -0.002
23 4 21.202 21.207 -0.005
24 1 52.858 52.814 0.25 m² 1 0.044
24 2 107.371 107.257 0.3 x 1.3m 0.114
24 3 26.219 26.222 -0.003
24 4 35.246 35.173 0.073
31 1 64.323 64.155 1.2 m² 1 0.168
31 2 111.203 111.178 1.15 x 1.3m 0.025
31 3 82.778 82.68 0.098
31 4 83.515 83.403 0.112
47 3 524.257 524.033 8.7m² 1 0.224
47 1 497.226 496.942 7 x 1.4m 0.284
47 2 128.589 128.611 -0.022
47 4 199.206 198.928 0.278
52 1 226.563 226.716 6 m² 1 -0.153
52 3 163.322 163.198 2.15 x 3.5m 0.124
52 4 181.653 181.874 -0.221
52 2 174.268 174.274 -0.006
147 4 565.224 565.044 6.5 m² 3 0.18
147 3 382.059 381.772 7 x 1.2m 0.287
147 2 137.639 137.472 0.167
147 1 365.551 365.506 0.045
536 4 98.998 99.111 11 m² 3 -0.113
536 3 836.468 836.576 10 x 1.1m -0.108
536 2 277.151 277.219 -0.068
536 1 345.192 345.244 -0.052
321 1 138.29 138.323 1 m² 1 -0.033
321 2 40.361 40.403 1.6 x 0,7 m -0.042
321 3 105.439 105.491 -0.052
321 4 83.741 83.761 -0.02
334 1 129.633 129.641 1.2m² 1 -0.008
334 2 60.907 60.932 1.6 x 0.8m -0.025
334 3 108.209 108.27 -0.061
334 4 80.202 80.215 -0.013
412 1 190.447 190.397 1.3m² 1 0.05
412 2 40.139 40.045 2.6 x 0.5m 0.094
412 3 137.58 137.604 -0.024
412 4 166.303 166.303 0
398 1 70.943 71.04 0.6m² 1 -0.097
398 2 58.798 58.751 0.9 x 0.8m 0.047
398 3 76.393 76.56 -0.167
398 4 48.634 48.576 0.058
554 1 115.586 115.568 1.2m² 1 0.018
554 2 65 65.012 1.2 x 1.3m -0.012
554 3 58.074 58.038 0.036
554 4 89.594 89.632 -0.038
10 1 139.878 139.69 1.66m² 1 0.188
10 2 59.091 58.918 0.9 x 1.8m 0.173
10 3 101.785 101.618 0.167
10 4 90.462 90.398 0.064
13 1 172.519 172.207 2.9m² 1 0.312
13 2 65.837 65.798 0.9 x2.29m 0.039
13 3 160.1 159.987 0.113
13 4 108.682 108.352 0.33
51 1 193.345 193.644 4.34m² 1 -0.299
51 2 187.316 187.082 2.1 x 2.4m 0.234
51 3 139.612 139.719 -0.107
51 4 89.815 90.003 -0.188
97 1 283.48 283.324 3.92m² 1 0.156
97 2 101.208 101.284 1.5 x 3.34m -0.076
97 3 156.24 156.252 -0.012
97 4 212.833 212.17 0.663
176 1 201.638 201.939 3.25m² 1 -0.301
176 2 101.637 101.564 1.34 x 2.4m 0.073
176 3 137.343 137.504 -0.161
176 4 72.319 72.273 0.046
47/95 1 641.496 641.418 8.9m² 2 0.078
47/95 2 105.265 105.423 6.98 x 1.28m -0.158
47/95 3 314.418 314.436 -0.018
47/95 4 379.341 379.324 0.017
147 1 627.263 627.529 6.37m² 3 -0.266
147 2 94.734 94.746 7 x 1.2m -0.012
147 3 296.632 296.837 -0.205
147 4 280.471 280.329 0.142
247 1 212.186 212.066 4.08m² 2 0.12
247 2 205.946 205.589 2.5 x 2.4m 0.357
247 3 197.994 198.416 -0.422
247 4 159.828 159.587 0.241
246 1 250.659 251.987 4.8m² 2 -1.328
246 2 208.503 208.436 3.1 x 2.2m 0.067
246 3 262.904 263.441 -0.537
246 4 158.483 159.26 -0.777
6 1 47.69 48.276 0.75 m² 1 -0.586
6 2 106.638 106.873 1.3 x 0.65 m -0.235
6 3 97.007 97.63 -0.623
6 4 92.938 92.535 0.403
12 1 117.916 117.911 0.668m² 1 0.005
12 2 26.988 26.698 0.46 x 1.5m 0.29
12 3 79.628 79.695 -0.067
12 4 68.86 68.566 0.294
11 1 138.311 138.297 0.712m² 1 0.014
11 2 45.143 45.175 0.54 x 1.75m -0.032
11 3 118.281 118.245 0.036
11 4 78.148 78.119 0.029
13 1 41.371 41.386 0.424m² 1 -0.015
13 2 39.893 39.872 0.47 x 0.52m 0.021
13 3 38.962 38.811 0.151
13 4 31.561 31.72 -0.159
15 1 122.66 122.627 0.831m² 1 0.033
15 2 46.402 46.337 1.27 x 0.68m 0.065
15 3 83.043 83.069 -0.026
15 4 63.647 63.633 0.014
20 1 132.39 132.401 0.798m² 1 -0.011
20 2 41.085 41.138 0.6 x 1.41m -0.053
20 3 93.914 94.003 -0.089
20 4 65.114 64.389 0.725
32 1 134.232 134.13 1.98m² 1 0.102
32 2 92.02 91.977 1.16 x 1.72m 0.043
32 3 123.213 123.185 0.028
32 4 102.014 101.938 0.076
5 1 91.013 90.945 0.391m² 1 0.068
5 2 36.363 35.739 1.08 x 0.39m 0.624
5 3 57.404 57.165 0.239
5 4 57.122 56.96 0.162
8 1 27.56 27.56 0.626 m² 1 0
8 3 44.699 44.627 0.43 x 1.18m 0.072
8 4 91.483 91.209 0.274
8 2 79.724 79.71 0.014

Fig. 3: Example profile. Red = Fotogrametric control points; Yellow = Endpoints of the measured axis

Results

The distances were measured in both programs and can thus be compared with one another. In order to quantify the deviation for each profile, the average of the amount of the deviation of each measured distance per profile was calculated and this was taken as an estimate. It was found that the mean deviation in all profiles is between 0.015 cm and 0.46 cm (mean 0.13 cm) (Fig. 4). The maximum deviations of the individual routes are between 0.03 and 1.33 cm with an average value of 0.28 cm (Fig. 5). The sizes of the profiles are between 0.2 - 11 m² and a correlation between profile size or length and errors cannot be determined. Overall, however, it can be said that even with the larger deviations of 1 cm, there may also be errors in the marking accuracy and these errors - especially in relation to hand drawings and taking line widths and measurement tolerances into account - are negligible.

Fig. 4: Box plot of the mean differences in the measured axis lengths

Fig. 4: Box plot of the mean differences in the measured axis lengths

Fig. 5: Box plot of the maximal differences in the measured axis lengths

Fig. 5: Box plot of the maximal differences in the measured axis lengths

Acknowledgement

We would like to thank the ISAAKiel working group, within the framework of which the plugin was created.

References

Backhaus, K., B. Erichson, W. Plinke, and R. Weiber. 2011.

Gütter, S. 2015. “Wie machen wir das eigentlich? – Ergebnisse unserer Umfrage zu den heute gebräuchlichen grabungstechnischen Dokumentationsmethoden.” Rundbrief Grabungstechnik 7: 1–19.

Mennenga, M., K. Schmütz, and C. Rinne. 2019. “ISAAKiel/profileAAR: ProfileAAR - First Official Qgis3 Release (Version 2.0.1).” GitHub Repository. GitHub. doi:https://doi.org/10.5281/zenodo.3234836.

QGIS Plugin Repository. 2021. “ProfileAAR.” https://plugins.qgis.org/plugins/profileAAR/.


  1. corresponding author

  2. co-first author